Distributed Contact Sensing Enabled by Vibration Propagation on Robot End-Effector

Abstract

To match the dexterity of human hands, a robot’s end-effector needs tactile sensing. However, current tactile sensing solutions often have complex electronics, making them impractical for covering the entire area of the manipulator without interfering with robot manipulation, especially for miniature objects. Here, we present a tactile sensing design that enables a single accelerometer positioned at the base of a robot’s end-effector, to locate and estimate contact forces applied across a large region of the end-effector. Inspired by human tactile sensing, where a single afferent responds to skin vibrations over a large area, we integrated a string to transmit vibrations from remote contacts to the accelerometer. We utilized lightweight machine learning models to decode tactile information from the vibration signals captured by the accelerometer. Our experimental results demonstrate that we can accurately predict remote contact locations and force amplitudes, with a precision of 1.9 mm and 0.09 N, respectively. Additionally, the vibrations can be used to identify the surface materials of contact objects, achieving 99% accuracy in discriminating between 15 different materials. Our approach could help simplify and minimize the design of robot manipulators, enabling more delicate manipulation and reducing the hardware costs and data volume required for tactile sensing.